Python 3.x 伯特在神经机器翻译模型中的应用 我正在开发一个英语-印地语神经机器翻译模型。我想使用BERT嵌入来生成单词向量，而不是Keras嵌入。我有什么办法可以做吗

我的编码器型号是

encoder_inputs = Input(shape=(None,))
enc_emb =  Embedding(num_encoder_tokens, latent_dim, mask_zero = True)(encoder_inputs)
encoder_CNN = Conv1D(16, kernel_size=11, activation='relu')(enc_emb)
encoder_lstm = LSTM(latent_dim, return_state=True)
encoder_outputs, state_h, state_c = encoder_lstm(encoder_CNN)
encoder_states = [state_h, state_c]

dec_emb_layer = Embedding(num_decoder_tokens, latent_dim, mask_zero = True)
dec_emb = dec_emb_layer(decoder_inputs)
decoder_CNN = Conv1D(16, kernel_size=16, activation='relu')(dec_emb)
decoder_lstm = LSTM(latent_dim , return_sequences=True, return_state=True)
decoder_outputs, _, _ = decoder_lstm(decoder_CNN, initial_state=encoder_states)
decoder_dense = Dense(num_decoder_tokens, activation='softmax')
decoder_outputs = decoder_dense(decoder_outputs)
model = Model([encoder_inputs, decoder_inputs], decoder_outputs)
model.summary()

我的解码器型号是

encoder_inputs = Input(shape=(None,))
enc_emb =  Embedding(num_encoder_tokens, latent_dim, mask_zero = True)(encoder_inputs)
encoder_CNN = Conv1D(16, kernel_size=11, activation='relu')(enc_emb)
encoder_lstm = LSTM(latent_dim, return_state=True)
encoder_outputs, state_h, state_c = encoder_lstm(encoder_CNN)
encoder_states = [state_h, state_c]

dec_emb_layer = Embedding(num_decoder_tokens, latent_dim, mask_zero = True)
dec_emb = dec_emb_layer(decoder_inputs)
decoder_CNN = Conv1D(16, kernel_size=16, activation='relu')(dec_emb)
decoder_lstm = LSTM(latent_dim , return_sequences=True, return_state=True)
decoder_outputs, _, _ = decoder_lstm(decoder_CNN, initial_state=encoder_states)
decoder_dense = Dense(num_decoder_tokens, activation='softmax')
decoder_outputs = decoder_dense(decoder_outputs)
model = Model([encoder_inputs, decoder_inputs], decoder_outputs)
model.summary()

感谢您的帮助，因为我在这个问题上遇到了麻烦